Hallin and Mancini Revisited

Brüggemann, Michael; Engesser, Sven; Büchel, Florin; Humprecht, Edda; Castro, Laia (2014): Hallin and Mancini Re- visited. Four Empirical Types of Western Media Systems. In Journal of Communication 64 (6), pp. 1037–1065. Avail- able online at https://doi.org/10.1111/jcom.12127.

Hallin and Mancini Revisited:

Four Empirical Types of Western Media Systems

Michael Brüggemann, Sven Engesser, Florin Büchel, Edda Humprecht, & Laia Castro*

University of Zurich

Abstract

The analysis of media systems has become a corner stone in the field of comparative communication research. Ten years after its publication, we revisit the landmark study in the field,

Hallin and Mancini’s “Comparing Media Systems” (2004), and operationalize its framework for standardized measurement. The study at hand is, to the best of our knowledge, the first to comprehensively validate the original dimensions and models using aggregated data from the same sample of Western countries. Three out of four dimensions of media systems show relatively high levels of internal consistency but “role of the state” should be disaggregated into three sub-dimensions. A cluster analysis reveals four empirical types of media systems that differentiate and extend the original typology.

Keywords: Media systems, international comparison, journalism, political communication, typology, cluster analysis

Acknowledgement: We would like to thank the head of our division Frank Esser and the

Institute of Mass Communication and Media Research for providing the freedom and the research environment that made this study possible. Also, we would like to thank the editors and anonymous reviewers of the journal for their highly instructive feedback.

Corresponding authors: Michael Brüggemann and Sven Engesser, IPMZ – Institute of

Mass Communication and Media Research, University of Zurich, Andreasstrasse 15, 8050

Zürich, Switzerland, e-mails: [email protected]; [email protected]

Hallin and Mancini Revisited 2

* The first two authors have equally shared the lead on this paper.

Hallin and Mancini Revisited 3

Hallin and Mancini Revisited:

Four Empirical Types of Western Media Systems

The study of media systems has become a corner stone in the evolving field of comparative communication research: typologies of media systems can serve as powerful heuristics that guide concept formation, hypotheses, and case selection. The unrivalled success of the two books, “Four Theories of the Press” (1956) and its successor “Comparing Media Sys- tems” (2004), proves the power of media typologies. Hallin and Mancini’s work has inspired and guided comparative research. It has received a lot of praise but also some criticism.

Ten years after its publication, it is time to revisit the theoretical framework and ad- dress a line of criticism that has been most prominently advanced by Norris (2011). She claims that Hallin and Mancini’s three models cannot be replicated due to a lack of precise operationalization and standardized measurement. This paper follows her call and contributes to the field by operationalizing, validating, and – in some respects – modifying the framework introduced by Hallin and Mancini. This allows for replication and extension in future studies.

The Challenge of Measuring Media Systems

Comparative research is about explaining the commonalities and differences of communication practices by looking at variation in their contexts (Blumler, McLeod, & Rosen- gren, 1992; Esser & Pfetsch 2004). Typologies of media systems have become powerful tools in this endeavor as they provide simplified models of these explanatory contexts. 1 They de- scribe typical patterns of how journalism cultures, media policy, media markets, and media use are connected in a given society. The idea of analyzing these patterns as “systems of political communication” was prominently advanced by Blumler and Gurevitch (1995, p. 12).

Today, Hallin and Mancini’s three models of media systems have become the point of reference for many comparative studies of journalism and political communication (e.g. Aal- berg & Curran 2012; Benson, Blach-Orsten, Powers, Willig, & Zambrano, 2012; Strömbäck,

Orsten, & Aalberg 2008; Esser et al. 2012; Voltmer, 2013). A number of authors have

Hallin and Mancini Revisited 4 discussed in what way countries not included in the original study fit into the framework and how it should be extended to other parts of the world (see the contributions in Hallin &

Mancini, 2012b). Hallin and Mancini (2004) openly acknowledge these limitations of their study. Their three models of media and politics are heuristics: simplified versions or ideal types of reality that real media systems will always diverge from. Also, they are aware of the fact that they do not cover countries beyond the Western world and neglect certain categories of comparison that might be decisive in media systems beyond Europe, such as media freedom and religion (Couldry, 2005; Engesser & Franzetti, 2011; Norris, 2009, p. 332). In fact,

Hallin and Mancini (2012a) themselves invite the research community to extend and modify their dimensions, indicators, and models.

The future study of media systems should indeed include new forms of digital communication and go beyond the narrow realm of traditional news production as demanded by some of the critics (Hardy, 2012; Norris, 2009). Yet before extending the framework to cover more countries and including additional variables, another criticism should be addressed more thoroughly than it has been done by previous research. Norris (2009, p. 334) asks for a rigor- ous empirical validation of Hallin and Mancini’s framework based on standardized indicators in order to see whether the dimensions “actually cluster together in meaningful ways.” In response to this, Hallin and Mancini (2012a) have stressed the exploratory nature of their framework that still “would need to be tested by empirical research” (p. 213).

However, Hallin and Mancini’s qualitative approach is not only a weakness but also strength of their study: It fosters an enhanced understanding of the cases under analysis. At the same time, thanks to the solid framework of dimensions provided, the explanations are not idiosyncratic but detect patterns of relations that go beyond understanding individual cases.

Yet a standardized approach can validate and complement Hallin and Mancini’s findings.

Our study is an attempt to perform the kind of operationalization that both Hallin and

Mancini (2004) and Norris (2009) encourage. We operationalize the framework in a

Hallin and Mancini Revisited 5 standardized way that validates Hallin and Mancini’s dimensions and models drawing on aggregated data – some of which was not yet available in 2004. The article will thus pursue the following three challenges and research questions related to them. The first question addresses the operationalization of Hallin and Mancini’s dimensions: How can they be measured drawing on standardized data (RQ1)? The second question explores the dimensions and their relations: Do the indicators combine to the dimensions ingrained in Hallin and Mancini’s framework (RQ2a) and do the dimensions correlate in the way the authors hypothesized (i.e. journalistic professionalism being positively linked to media-market development and political parallelism linked to state intervention) (RQ2b)? The third question is dedicated to Hallin and

Mancini’s models: Do the countries actually cluster into empirical types that can be characterized along the lines of the three models (RQ3)?

Revisiting the Framework of “Comparing Media Systems”

Hallin and Mancini (2004) have substantially advanced comparative research with three contributions. Their analytic framework identifies: (1a) basic dimensions for the analysis of media systems; (1b) dimensions of political systems that shape media systems; and (2) the resulting three models of media systems.

These contributions, however, have received uneven attention. While the models have become the reference point for various comparative analyses, even for those which do not deal with political communication, the analytic framework that generated these models seems to be somewhat neglected. This is unfortunate because the dimensions “travel better” and are more easily applied to other countries (Hallin and Mancini, 2012b, p. 287) than the models, which derive from a limited number of case studies on Western countries only.

This paper, in line with the main thrust of Hallin and Mancini’s book, focuses on the identification of dimensions and types of media systems. Future studies, however, should also give closer scrutiny to the political system variables.

Dimensions of Media Systems

Hallin and Mancini Revisited 6

Therefore, the dimensions of “Comparing Media Systems” deserve closer attention before moving on to the well-known three models. Hallin and Mancini (2004) introduce four dimensions for the analysis of media systems and they also discuss several indicators that can be subsumed under each dimension. Hallin and Mancini (2004) use the term “dimensions” in a relatively broad way. However, a standardized measurement of dimensions, as well as an assessment of their internal consistency and correlations requires a stricter definition: Dimen- sions have to vary on a continuum between two poles, such as less or more political parallelism. While two of the original dimensions are indeed dimensions in this narrower sense of the term (i.e. political parallelism and journalistic professionalism), the other two (i.e. media market and role of the state) are rather not.

On the one hand, we suggest relabeling the dimension of media market. Instead of comprehensively discussing “media markets”, Hallin and Mancini focus on the history and current state of what we prefer to entitle inclusiveness of the press market. This determines whether the press only reaches out to the elites or to a broader mass audience. On the other hand, as the fourth dimension role of the state is much broader than the other ones, we distinguish three sub-dimensions, each one built on a different form of state intervention. Overall, we propose the following conceptualization (for the details see Table 3):

Inclusiveness of the press market. This dimension denotes how far the press reaches out to a broader audience. It includes indicators like the general reach of newspapers, as well as, more specifically, the reach among women and men, or among different segments of society, such as the working class (Hallin & Mancini, 2004, pp. 22-26).

Political parallelism. This dimension measures to what extent political advocacy is conceived of being part of the mission of journalism in different countries. It extends the concept of party-press parallelism that describes the links between media outlets and political parties (Blumler & Gurevitch, 1995; Seymour-Ure, 1974) to include general political values and being close to certain political camps rather than to parties (Hallin & Mancini, 2004, p. 28).

Hallin and Mancini Revisited 7

The dimension encompasses indicators such as the extent to which media coverage is shaped by journalists’ political affiliations, the degree to which audiences consume media according to their political preferences, the separation between news and commentary, political bias in news reporting, and the public broadcasters’ dependence on the government (Hallin &

Mancini, 2004, pp. 28-31).

Journalistic professionalism. Hallin and Mancini (2004, pp. 34-36) discuss three indicators for journalistic professionalism. The first is the degree of professional autonomy that journalists enjoy as a group. Autonomy can be limited either by external forces, such as political or economic actors, or by actors within the news organizations, such as the publishers or the owners. The second indicator for journalistic professionalism is the development of distinct professional norms, such as common ethical principles, e.g. concerning the protection of confidential sources. The third indicator is the extent to which journalists are oriented towards an ethic of serving the public interest. The absence of journalistic professionalism manifests itself in the instrumentalization of journalists by economic or political interests, which in turn contribute to diminish their credibility.

Role of the state. The role of the state differs between countries as far as the extent and the direction of state interventionism is concerned (Hallin & Mancini 2004, p. 41). As the form of state intervention and not only the degree of state interventionism might vary, it is not really one dimension of a media system but a broader multi-dimensional category. One may distinguish state interventionism that complements private media by public media, measures of the state that support private media, and measures that restrict media. Following this line of reasoning, we suggest introducing three dimensions that help to characterize Western media systems and are also open to standardized measurement.

In Western countries, public broadcasting is the most important kind of state intervention, following Hallin and Mancini (2004, p. 41) but also Curran et al. (2009) and Aalberg &

Curran (2012). Western countries vary widely on this dimension, with the US maintaining

Hallin and Mancini Revisited 8 only marginal public broadcasting and Germany providing billions of Euros for a broad vari- ety of national and regional public media outlets.

The second dimension of state intervention is support for commercial media organizations in the form of direct and indirect press subsidies. Some countries have strong public broadcasting outlets but not necessarily subsidies for the press. We are therefore dealing with two distinct dimensions of a media system.

The third dimension covers state interventions that restrict media organizations. This may happen by constraining media content or by regulating the media market. Western countries have abolished open censorship and guarantee press freedom. They limit media content in comparatively modest ways, for example with libel laws, more or less restricted access to public information and also by regulating election campaign coverage (Hallin & Mancini,

2004, p. 43-44). These measures are very diverse and can, from a theoretical point of view, hardly be integrated into a single index. While they should be explored further in future comparative research, the questions of regulating media ownership seems to be a more fundamen- tal variable distinguishing Western media systems (Baker, 2007). It should be stressed, however, that limits on news content and press freedom are probably the most important dimension to distinguish media systems on a global scale.

We do thus propose to disaggregate role of the state into three dimensions relating to public broadcasting, press subsidies, and media ownership regulation. These issues are all included in Hallin and Mancini’s role of the state dimension and they are all highly relevant in shaping media systems. They are also conceptually distinct so that one should not aggregate them into a single dimension before empirically exploring their relation.

The key interest behind Hallin and Mancini’s (2004) quest for models of media systems was “the identification of characteristic patterns of relationship between system characteristics” (p. 11). If these patterns were absent, there would be no such thing as a media system while the existence of these patterns would be an indicator of ‘systemness’. They would

Hallin and Mancini Revisited 9 show us that if we modified one element of the system, other elements would be likely to change as well (Engesser & Franzetti, 2011).

Therefore identifying statistical correlations would strengthen our claim that the dimensions introduced above can indeed be regarded as constituting a media system. Having said that, Hallin and Mancini (2004, p. 45) also claim that their dimensions are “ultimately ir- reducible to one another”. Empirically, this could be measured by looking at the strength of the correlations: if the different dimensions were to perfectly correlate, this would question their value as separate dimensions of a media system.

Hallin and Mancini (2004, pp. 38, 39, 300, 301) also formulate hypotheses on how the dimensions are interrelated. They reckon that political parallelism is to some degree negatively related to journalistic professionalism because “[…] historically, the development of journalistic professionalization eroded political parallelism” (p. 38). Professionalism is expected to be positively related to the development of a strong commercial press. Newspapers that do not depend on financial support from the state are more likely to follow their own professional logic. Also, high degrees of political parallelism and strong state intervention are linked to some degree: influential parties and a strong role of the state may go hand in hand

(p. 300).

Three Models of Media and Politics

Hallin and Mancini’ (2004) analysis results in a typology of three models of media and politics that will briefly be summarized in the following (see also Table 1):

The North Atlantic or Liberal model. This model comprises Great Britain, the USA,

Canada, and Ireland, and is characterized by high reach of the press market, low degrees of political parallelism, a highly professionalized journalism and a weak role of the state.

The Northern European or Democratic Corporatist model. It comprises the Nordic countries, the German speaking countries, Belgium, and the Netherlands. This model displays a high reach of the press market, relatively high degrees of political parallelism, strong

Hallin and Mancini Revisited 10 professionalization and strong state intervention, in the form of strong public service broadcasters and subsidies for the press.

The Mediterranean or Polarized Pluralist model. France, Greece, Italy, Portugal, and Spain form this group that is characterized by a low reach of the daily press, high political parallelism, weak professionalization, and strong state intervention. The latter does not necessarily mean that the state effectively serves the public interest. Particularistic interests and cli- entelistic relationships can also lead to failed state intervention (Hallin & Mancini 2004, p.

138). Besides, some of the Polarized Pluralist countries display attempts of “savage” deregu- lation in the broadcast sector (Hallin & Mancini, 2004, p. 73).

[Table 1 about here]

There have been some issues with the classification of particular countries within this framework. Most intensively discussed is the case of Great Britain with its strong public broadcasting, its ideologically polarized press, and its lack of fit with the Liberal model

(Humphreys, 2012; Norris, 2009, p. 334). In line with their critics, Hallin and Mancini (2004) point out: “The common idea of an ‘Anglo-American’ model of journalism is in part a myth”

(p. 69). Curran (2010) stresses the exceptionalism of the US, in particular with regards to its role in world politics and the ‘semi-independent’ role of the press regarding the public justifi- cation of military interventions. Hardy (2008) even considers opening up a fourth separate model for the US. Another interesting case is Portugal which has been considered to “diverge significantly” from the Polarized Pluralist model and to “move away” from the political parallelism that characterizes Spain, Italy, and Greece (Hallin & Mancini, 2012a, p. 209).

Finally, the case of Germany may be interesting. With its absence of direct press subsidies it does not fit with the Nordic countries (Humphreys 2012, p. 163). Rather, it is somewhat similar to the British case with its very strong public broadcasting, high levels of

Hallin and Mancini Revisited 11 political parallelism in the press and an absence of press subsidies. Thus, some countries’ classifications in the framework are particularly contested and deserve critical review.

Method

Hallin & Mancini’s (2004) dimensions and models were examined using publicly available data and common statistical methods (e.g. correlation and cluster analysis).2 Alt- hough this approach admittedly risks some oversimplification, we believe that it facilitates replication by condensing hazy relations between theoretical dimensions and media systems into solid values that can be used in subsequent studies.

The utilized datasets may be partially incomplete, outdated, or inequivalent. We ac- commodate for these problems by complementing and cross-validating the existing datasets.

We drew on studies that employed a multitude of methods, such as expert interviews, surveys, literature reviews, and content analyses. We also took care that these studies were as up-to- date as possible and that they referred to the same period of time (i.e. from 2007 to 2011). All data sources are briefly portrayed in Table 2.

[Table 2 about here]

Operationalization and Data

The biggest challenge of this study was operationalizing the six dimensions (RQ1). In order to base our measurement on a solid foundation, every dimension was, first of all, divided into several indicators (see Table 3). For some dimensions (i.e. political parallelism and journalistic professionalism), the available data allowed us to use five or six indicators, for other dimensions (i.e. public broadcasting and press subsidies) we had to limit ourselves to two indicators. In any case, the indicators complemented and validated each other.

[Table 3 about here]

Hallin and Mancini Revisited 12

For the dimension of inclusiveness of the press market we obtained most of the required data from the World Press Trends (WPT) (2008-10). In the case of the indicator working class daily newspaper reach we had to rely on two different sources: For roughly half of the countries, we used the World Values Survey (WVS) (2005-07), and for the other half we drew on the European Election Studies (2009).

Concerning the dimensions of political parallelism and journalistic professionalism, the measures for two countries differed from the rest. For all the EU-member states in our sample and Norway, we found the required data in Popescu et al.’s (2011) European Media

Systems Survey (EMSS). As Switzerland and the USA were not included in this study, we switched to the Worlds of Journalism Survey (WJS) (2007-11) for these countries. From this source, we selected those questions which we considered equivalent to the ones from EMSS

(2010). In order to allow for comparison, the values for these questions, originally measured on the five-point scale of WJS, were normalized to the eleven-point scale of EMSS. In order to analyze the empirical equivalence of the two datasets we calculated bivariate correlations between the interchanged questions which resulted in acceptable coefficients (r = .70 – .97).

In terms of political parallelism, we could draw on the indices introduced by Popescu et al. (2011), Hanretty (2009), and van Kempen (2007). This broadened the basis of our measurement further and also increased its intersubjectivity. Concerning the indices by Popescu et al. (2011) and Hanretty (2009) we used the data provided by the authors. For the index by van

Kempen (2007), we did not take the original data from the European Election Studies (1999) but reran the calculations on an updated version of the dataset from 2009.

Since Hanretty’s (2009) analysis of public broadcasting independence did not include

Greece and the Netherlands, we substituted the missing values with information from Hum- phreys (1996) on the politicization of public broadcasting systems. Again, a strong correlation between the two datasets (r = .75) shows that they are, at least to a certain degree, equivalent.

Hallin and Mancini Revisited 13

For the dimensions public broadcasting, ownership regulation and press subsidies we used data from the European Audiovisual Observatory (EAO) (2011) and, again, from the

World Press Trends (WPT) (2008-10).

The indicator values were, if they had been measured on different scales, z-standardized. Subsequently they were averaged to dimension indices of acceptable internal consistency (Cronbach’s α = .59 – .91).3 We preferred average to additive indices because they are less sensitive to missing values, which could, in spite of complementary datasets, not be completely avoided. Prior to data analysis, the dimension indices were also z-standardized.

Country Sample and Data Analysis

For comparative reasons, we intended to apply the instrument to the same countries as

Hallin and Mancini (2004) (see Table 1). Unfortunately, Canada was covered by neither

EMSS (2010) nor WJS (2007-11), so we had to leave it out of our study. This resulted in a total sample of 17 Western countries.

We explored the relations between the six dimensions by conducting bivariate correlation analyses between the z-standardized dimension indices. For validating Hallin and

Mancini’s (2004) models of media systems we used the six z-standardized dimension indices to carry out a two-stage cluster analysis of the seventeen countries. In order to identify the number of clusters we performed a hierarchical cluster analysis using Ward’s algorithm and the Squared Euclidean distance as heterogeneity measure. We chose the four-cluster solution for three reasons. First, merging the clusters beyond the fourth would result in solutions that were too heterogeneous (see Table 4). If we display the sum of squared distances as a scree plot, this is reflected by a strong elbow at the fourth cluster. Second, the dendogram for the four-cluster solution is very clear and highly interpretable (see Figure 1). Third, we checked the clarity and interpretability of alternative solutions and found that they could not compete with the four-cluster solution.

Hallin and Mancini Revisited 14

[Table 4 and Figure 1 about here]

In order to optimize the countries’ cluster membership and validate the analysis through another algorithm we used the centroid-based k-means method (Milligan & Sokal,

1980). By and large, the cluster solution was confirmed: only France shifted from the Central to the Southern cluster. This change appears plausible because the Southern cluster contains those countries which were labeled Polarized Pluralist by Hallin and Mancini (2004).

Findings

The findings of this study are organized in two sections: the first assumes a variable- oriented perspective and is dedicated to the internal consistency and the correlations of the six dimensions of media systems (RQ2a and RQ2b); while the second is case-oriented and pre- sents the results of the cluster analysis carried out to empirically validate Hallin and

Mancini’s models of media and politics (RQ3).

Consistency and Correlations of the Dimensions

The three dimensions inclusiveness of the press market, political parallelism, and journalistic professionalism showed acceptable levels of internal consistency (see Method section). This can be regarded as empirical support for Hallin and Mancini’s original conceptualization, and thus we retained these dimensions as they were.

However, role of the state proved to be a multi-dimensional category. A superordinate dimension of state interventionism combining the different indicators of public broadcasting, ownership regulation, and press subsidies would be of unacceptable internal consistency

(Cronbach’s α = .36). This shows that role of the state is a category consisting of different dimensions, as we already assumed theoretically.

In order to identify these dimensions, a Principal Component Analysis (PCA) of the different indicators for role of the state was conducted. It revealed three components with Ei- genvalues of λ1 = 2.18, λ2 = 1.53, and λ3 = 1.34, explaining altogether 72 % of total variance.

Hallin and Mancini Revisited 15

Since a hypothetical fourth component would only have an Eigenvalue of λ4 = 0.79, the Kai- ser criterion suggested the extraction of three components. If we display the Eigenvalues as a scree plot, there is a strong elbow at the fourth component, also supporting the three-component solution. The indicators for ownership regulation load highly on the first component, the second component includes the indicators for press subsidies, and the indicators for public broadcasting correlate with the third component (see Table 5). Even though the results of this

PCA have to be treated with caution due to the small sample size and the limited number of variables loading on the components, the three extracted components, by and large, can be considered as empirical manifestations of role of the state’s three dimensions as introduced in the theoretical part of this paper.

[Table 5 about here]

Hallin and Mancini (2004) argue that a strong press market and highly professional journalism go hand in hand. This correlation is supported by our empirical analysis (see Table

6). They also assume that political parallelism and an interventionist state presuppose each other. However, we only find a relatively weak and non-significant positive correlation between political parallelism and one dimension of role of the state, which is ownership regulation. The correlations between political parallelism and the other two dimensions of role of the state (i.e. public broadcasting and press subsidies) are also non-significant, and even nega- tive. These empirical results show again that the role of the state proves to be more complex than expected and should be divided into three dimensions. The supportive form of state interventionism, as expressed by press subsidies, and the restrictive approach manifested in ownership regulations, correlate negatively, although non-significantly.

[Table 6 about here]

Hallin and Mancini Revisited 16

We also found that the stronger the political parallelism is in a country, the less professional are its journalists and the less inclusive is its press market. There is also evidence that an inclusive press market and a strong public broadcasting system co-occur.

Four Empirical Types of Western Media Systems

In the following section, we will present the results of the cluster analysis. The emerg- ing clusters can be regarded as empirical types of Western media systems which will be con- trasted to Hallin and Mancini’s more ideal-type models of media systems (see Table 7).

Grouping the seventeen Western countries in our sample together according to their position on the six dimensions results in four clusters: Central, Northern, Southern, and West- ern. We named them according to the geographic location of the countries they include.

[Table 7 about here]

The countries which originally formed the Democratic Corporatist model are now dis- tributed among the Northern, Central, and Western cluster. The relatively high homogeneity of the Nordic countries Denmark, Finland, Norway, and Sweden justifies the formation of a separate Northern cluster. Norway has the shortest distance to the center of this cluster and can be regarded as its prototype. Austria, Germany, and Switzerland constitute the Central cluster where they are joined by Great Britain. Germany comes closest to the center of this cluster and can be considered prototypical. Belgium and the Netherlands are detached from the Democratic Corporatist countries. They are more similar to the Liberal countries Ireland and the USA in the Western cluster. Most of the countries from the Polarized Pluralist model are absorbed by the Southern cluster, of which Italy is the prototype. Only Portugal does not fit in and finds its way to the Western cluster, for which it can be regarded prototypical.

Hallin and Mancini Revisited 17

Great Britain’s detachment from the Liberal countries and Portugal’s move away from the Polarized Pluralist states is in accordance with the theoretical considerations expressed by

Humphreys (2012), Norris (2009), as well as Hallin and Mancini (2012a) themselves. Great

Britain is less liberal and Portugal more liberal than originally conceptualized. The latter also applies to Belgium and the Netherlands. Portugal, Belgium, and the Netherlands share a non- interventionist state which manifests most clearly in relatively weak public broadcasting.

Therefore, they are transferred to Ireland and the USA in the Western cluster.

If we look at the cluster profiles (see Figure 2) we can see that the Central cluster is mainly characterized by strong public broadcasting, strict ownership regulation, and low press subsidies. The Northern countries show highly professional journalism, an inclusive press market, powerful public broadcasting, and generous press subsidies. This goes together with the lowest levels of ownership regulation and political parallelism among the four clusters. At this point, it becomes evident that one of the advantages of dividing the role of the state into three dimensions is to differentiate between Central and Northern countries. The Southern type combines the highest degree of political parallelism with the least professional journalism and the least inclusive press market. It is relatively heterogeneous in terms of ownership regulation. Countries from the Western type share a very low level of public broadcasting and press subsidies, both of which are exceptionally low in case of the USA.

[Figure 2 about here]

A basic measure of intra-cluster homogeneity can be obtained by dividing the standard deviation of each cluster variable within the cluster by the standard deviation of the cluster variable across the entire sample. If this coefficient exceeds 1, the intracluster standard deviation is larger than the total standard deviation and the cluster is considered heterogeneous. If the value scores substantially below 1, the cluster is relatively homogenous. In this paper, the

Hallin and Mancini Revisited 18 average values of the clusters across all six dimensions range from .51 (Central) to .73 (South- ern) which indicates satisfactory levels of homogeneity.

The empirical types presented in this section, by and large, correspond to Hallin &

Mancini’s (2004) characterization of their original models of media and politics (Table 1).

There are, however, some important differences. The political parallelism of Central and, most of all, Northern countries is lower than predicted by Hallin and Mancini (2004) for the

Democratic Corporatist model. Besides, the state interventionism of the Central and Northern type is not equally strong on all three dimensions: Central countries avoid granting press subsidies while Northern countries refrain from ownership regulation. Furthermore, in Southern countries, the state is not as much interventionist as attributed to the Polarized Pluralist model by Hallin and Mancini.

The ideal-type Liberal model with a highly inclusive press market and very professional journalists does not exist in the sample. Even the presumed prototype USA does not come up with high levels on these dimensions. Instead we find the Western type as empirical manifestation of the Liberal model: Countries of this type share low levels of state intervention but only medium levels of press market and journalistic professionalism.

As final step of the data analysis we investigated how well the four empirical types explained the variance of the six dimensions of media systems (see Table 8). The explanatory power of the clusters is statistically significant across five dimensions and tends to be significant for the dimension of press subsidies. Our empirical types explain, on average, 62 % of variance. We can now compare the empirical types to Hallin & Mancini’s (2004) models:

Even though the original models already perform very well and explain a third of the variance, the empirical types increase this value by another 29 percentage points.

[Table 8 about here]

Hallin and Mancini Revisited 19

Discussion

The first challenge was to operationalize Hallin and Mancini’s theoretical framework for standardized measurement (RQ1) which was in fact feasible by drawing on standardized measurement and statistical analysis.

The second challenge was to explore the dimensions ingrained in Hallin and Mancini’s framework and their empirical relations (RQ2a and RQ2b). The dimensions inclusiveness of the press market, political parallelism, and journalistic professionalism displayed high degrees of internal consistency and may thus serve as the base for future studies. In line with the hypotheses from Hallin and Mancini (2004) we find that journalistic professionalism goes hand in hand with a strong and highly inclusive press market, while it is negatively correlated with political parallelism.

Conceptually, we have argued that role of the state is not a one-dimensional but at least a three-dimensional category. The empirical analysis confirms our argument. A PCA of the indicators reveals that there are three sub-dimensions (public broadcasting, ownership regulation, press subsidies). Looking at specific cases illustrates why differentiating the dimension role of the state is helpful: Germany and Great Britain display high levels of state intervention with regards to ownership regulation and broadcasting but low levels in terms of press subsidies. These are also important features that distinguish Great Britain from the USA, and prevent us from putting both countries into the same group. This point has been raised by many of Hallin and Mancini’s critics (Curran 2011; Humphreys 2012; Norris 2009) and by themselves (Hallin & Mancini, 2004, p. 11). Empirically, the contributors to Aalberg and Cur- ran (2012) have demonstrated how the UK is much closer to other European countries with regard to a number of variables, particularly the powerful effect of the strong British public broadcasting on the political knowledge of the citizenry.

Thus, the empirical data confirms our theoretical proposition to distinguish three dimensions that might also be useful for future modeling of media policy: the state may support

Hallin and Mancini Revisited 20 the press with subsidies. It may furthermore complement commercial media by introducing and sustaining a strong public broadcasting system. Finally, policy makers may also restrict the free play of the market forces by imposing ownership regulations.

The third challenge was to determine empirical types of media systems and compare them with Hallin and Mancini’s three models (RQ3). We arrived at four types but there is a great overlap with the original typology: Our Southern type is very similar to the Polarized

Pluralist model and the Western type is close to the Liberal model.

Yet, we do also show, where the traditional models can no longer account for today’s media cultures. Three models may be a more parsimonious solution than four, yet, across all the dimensions the empirical types explain almost a third of variance more than the original models. The additional explanatory power not only applies to the new dimensions introduced into the framework, but also to all the dimensions that have been left unchanged. The better fit of the four types with the empirical data may also be expressed in very concrete terms, when looking at the affiliations of the countries: e.g. Great Britain’s detachment from the type of

Liberal media system as represented by the US. Also, Portugal’s move away from the Polar- ized Pluralist states is in line with claims raised in response to Hallin and Mancini’s study by

Humphreys (2012), Norris (2009), as well as Hallin and Mancini (2012a) themselves.

The empirical types found in this study challenge the three ideal-typical models of media systems in two ways: The first concerns the conceptualization of the Liberal model and the second implies a differentiation of the Democratic Corporatist model.

We suggest that the Liberal model needs to be refined when being confronted with the more recent empirical data included in this study: While historically, a mass-market press and journalistic professionalism entailed low levels of state intervention, today, there is a group of

“liberal” countries (our Western type) that share low levels of state involvement in the media sphere but they are no longer the front-runners in terms of inclusiveness of the press market and journalistic professionalism. This role is taken over by countries that subsidize the press,

Hallin and Mancini Revisited 21 display high levels of journalistic professionalism and low level of political parallelism (our

Northern type). Also, our correlational analysis does not support the expectation attached to the Liberal Model that low state interventionism in terms of press subsidies and public broadcasting entails high inclusiveness of the press market and journalistic professionalism.

The empirical types of media systems diverge in a second way from the original models. The Democratic Corporatist model corresponds to two empirical types: the Northern type, consisting of the Nordic countries, and the Central type, represented by cases from Central

Europe and Great Britain. Both types display similar levels of press market inclusiveness and a strong public broadcasting, but they do also differ in important ways. The Central type has lower levels of press subsidies and journalistic professionalism. It also displays higher levels of ownership regulation and political parallelism than the Northern type.

The four empirical types seem to provide a better fit with today’s media cultures: The

Nordic countries are not grouped together with countries like Germany, Austria and Britain.

The latter are characterized by a press that displays substantial political parallelism while the former all range at the lower end of the parallelism scale (also see Norris, 2011, p. 257). The

Democratic Corporatist model of Hallin and Mancini (2004) did include many more countries than the other models; it has been criticized as some kind of residual category between the ex- treme points marked by their Liberal and their Polarized Pluralist models.

Introducing a fourth type also catches up with current conceptualizations and research practices in the field of international communication. Curran, Iyengar, Lund, and Salovaara-

Moring (2009) distinguish between three systems: “public service”, “dual”, and “market”.

They assign a Nordic country as representative of the former, Great Britain as representative of the dual model, and the USA as representative of the market model. It seems to make sense for many researchers not to group Britain with the US and also distinguish it from the North- ern media systems. When looking at how researchers try to represent the diversity of

Hallin and Mancini Revisited 22

European media systems in their studies, they often sample a Nordic country, a country from central Europe, a Southern, and a Western country (e.g. van Dalen, de Vreese, & Albæk,

2012).

Differences between the models and country classification as presented by Hallin and

Mancini (2004) can be explained by the different approaches but also by the time frames of the two studies: Hallin and Mancini could not draw on the more comprehensive and up-to- date data sets included in this study, but they have included a process perspective looking at the historical roots of different media systems and their evolution over time. The empirical types found in this study represent the status quo while the original three models reflect long- term historical developments. The fact that this study confirms many of the patterns from the three models shows that media systems are remarkably stable and that the models of Hallin and Mancini, by and large, reflect empirical reality.

Our empirical types may thus inspire future attempts to refine the ideal-typical formu- lations of models of media systems. Going beyond correlations between single variables they continue to serve as important steps towards developing a theory that grasps systematic patterns of relationships in political communications.

Limitations and Outlook

Our study confirms many of the assumptions and observations advanced by Hallin and

Mancini (2004), and it is able to further develop their framework in ways that may inspire future research. Yet our study has a number of limitations, mostly related to the data available to conduct this kind of analysis but also related to its quantitative approach that needs to be complemented by more in-depth case studies.

One limit is its reliance on expert surveys like the European Media Systems Survey

(EMSS) which can be criticized as not providing the precision of other data gathering methods because they introduce a certain degree of subjectivity. However, Popescu et al. (2010) report high reliability scores and support the validity of their results by comparing them with

Hallin and Mancini Revisited 23 those obtained by other studies. Besides country experts provide qualitative assessments which can be a valuable complement of standardized analysis.

Some problems also arise from the use of the World Press Trends (WPT) compilation.

While being a valuable resource, it is not a scientific study and has not been subjected to the kind of validity and reliability tests that should be applied to scientific cross-national studies.

In particular our binary indicators of state intervention drawn from WPT should be refined as soon as more comprehensive data is available.

These limitations notwithstanding, we do encourage researchers to continue to work with the available data, especially as we find that data from the different sources correlates highly on many indicators under analysis. However, we have to be cautious not to over-inter- pret data on single countries. For an in-depth understanding of individual cases qualitative analyses remain to be the superior research strategy.

Altogether, the results of the current study contribute to our knowledge of media systems by operationalizing, validating, and complementing Hallin and Mancini’s framework. It offers links and directions for further research in four respects.

First, this study was focused on the media systems variables introduced by Hallin and

Mancini (2004). As pointed out above their book also offers a discussion of related political systems. Following the formula provided by this study, future work should operationalize characteristics of the political system and explore their relation to the media-system dimensions analyzed in our study.

Second, the dimensions and indicators introduced above may be used in future studies as independent variables. The countries under analysis are all assigned with a numerical value for every dimension (see Table 9). These values may be used to explain variations in, for in- stance, media content or journalistic attitudes and practices. If these dependent variables also reach a metric level of measurement, correlative analyses become possible.

Hallin and Mancini Revisited 24

Third, the four types may help to provide a first rough guide in case selection and in formulating hypotheses in comparative studies of Western media systems. As a caveat, one should, however, reiterate the limits of our study that are inherited from Hallin and Mancini as discussed above: we do not cover all dimensions that might be relevant for an analysis of media systems, especially beyond the scope of political communication, beyond the traditional media and beyond Western countries.

This leads to the fifth and major challenge for research: going beyond Hallin and

Mancini (2004) rather than only revisiting it. Future studies should include additional dimensions into the analysis such as internet access, social media, and press freedom. This would also allow for easier application of the framework to non-Western countries. An impediment is the lack of data even for very important countries such as Brazil, Russia, India, and China.

A first and easier step would be to include Eastern European countries, for which the same kind of data is available that we used in our study.

Finally, the challenge remains to map the development of media systems over time:

This remains one of the domains where qualitative in-depth case studies of single or small numbers of countries should be conducted. Especially those countries who are in the process of shifting their alignment to another type of media system should be of special interest for this kind of study that could explain the how and why of these shifts. The indicators and indices developed and validated in this study may hopefully serve as a small step towards engag- ing with this wider agenda of comparative research.

Hallin and Mancini Revisited 25

References

Aalberg, T. & Curran, J. (Eds.). (2012). How media inform democracy: A comparative ap-

proach. New York: Routledge.

Bailey, K. D. 1994. Typologies and taxonomies: An introduction to classification techniques.

London: Sage.

Baker, C. E. (2007): Media concentration and democracy. Why ownership matters. New

York: Cambridge University Press.

Benson, R., Blach-Orsten, M., Powers, M., Willig, I., & Zambrano, S. V. (2012). Media sys-

tems online and off: Comparing the form of news in the United States, Denmark, and

France. Journal of Communication, 62(1).

Blumler, J. G., & Gurevitch, M. (1995). The crisis of public communication. London:

Routledge.

Blumler, J. G., McLeod, J. M., & Rosengren, K. E. (1992). An introduction to comparative

communication research. In J. G. Blumler, J. M. McLeod, & K. E. Rosengren (Eds.),

Comparatively speaking: Communication and culture across space and time (pp. 3–

18). Newbury Park: Sage.

Couldry, N. (2005). Review of comparing media systems. Political Studies Review, 3(2), 308.

Curran, J., Iyengar, S., Brink Lund, A., & Salovaara-Moring, I. (2009). Media system, public

knowledge and democracy: A comparative study. European Journal of Communica-

tion, 24(1), 5–26. DOI: 10.1177/0267323108098943.

Curran, J. (2011). Questioning a new orthodoxy. In J. Curran (Ed.), Communication and soci-

ety: Media and democracy (pp. 28–46). London: Routledge.

DeVellis, R. (2003). Scale development: Theory and applications. Thousand Oaks: Sage

Engesser, S., & Franzetti, A. (2011). Media systems and political systems: Dimensions of

comparison. International Communication Gazette, 73(4), 273–301. DOI:

10.1177/1748048511398590.

Hallin and Mancini Revisited 26

Esser, F. & Pfetsch, B. (2004). Meeting the challenges of global communication and political

integration: The significance of comparative research in a changing world. In F. Esser

& B. Pfetsch (Eds.), Comparing political communication: Theories, cases, and chal-

lenges (pp. 384–411). Cambridge: Cambridge University Press.

Esser, F., de Vreese, C. H., Strömbäck, J., van Aelst, P., Aalberg, T., Stanyer, J., et al. (2012).

Political Information Opportunities in Europe. The International Journal of Press/Pol-

itics, 17(3), 247–274.

Hair, J. F, Anderson, R. E., Tatham, R. L., & Black, W. C. (1998). Multivariate data analysis.

Englewood: Prentice Hall International

Hallin, D. C., & Mancini, P. (2004). Comparing media systems: Three models of media and

politics. Cambridge: Cambridge University Press.

Hallin, D. C., & Mancini, P. (2012a). Comparing media systems: A response to critics. In F.

Esser & T. Hanitzsch (Eds.), Handbook of comparative communication research

(pp. 207–220). London: Routledge.

Hallin, D. C.; Mancini, P. (Eds.) (2012b): Comparing media systems beyond the western

world. Cambridge: Cambridge University Press.

Hanretty, C. (2009). Explaining the de facto independence of public broadcasters. British

Journal of Political Science, 40, 75–89. DOI: 10.1017/S000712340999024X.

Hardy, J. (2008). Western media systems. New York: Routledge.

Hardy, J. (2012). Comparing media systems. In F. Esser & T. Hanitzsch (Eds.), Handbook of

comparative communication research (pp. 185–206). London: Routledge.

Humphreys, P. (1996). Mass media and media policy in Western Europe. Manchester: Man-

chester University Press.

Humphreys, P. (2012). A political scientist’s contribution to the comparative study of media

systems in Europe: A response to Hallin and Mancini. In N. Just & M. Puppis (Eds.),

Hallin and Mancini Revisited 27

Trends in communication policy research: New theories, methods and subjects

(pp. 141–158). Bristol, UK: Intellect.

Milligan, G. W., & Sokal, L. M. (1980). A two-stage clustering algorithm with robust recov-

ery characteristics. Educational and Psychological Measurement, 40, 755-759.

Norris, P. (2009). Comparative political communications: Common frameworks or Babelian

confusion? Government and Opposition, 44(3), 321–340. DOI:

http://dx.doi.org/10.1111/j.1477-7053.2009.01290.x

Norris, P. (2011). Comparative communications. In D. Caramani (Ed.), Comparative politics

(pp. 352–370). Oxford: Oxford University Press.

Pfetsch, B., & Esser, F. (2012). Comparing political communication. In F. Esser & T. Ha-

nitzsch (Eds.), Handbook of Comparative Communication Research (pp. 25–47). Lon-

don: Routledge.

Popescu, M., Toka, G., Gosselin, T., & Santana Pereira, J. (2011). European media systems

survey 2010: Results and documentation. Colchester: University of Essex.

Seymour-Ure, C. (1974). The political impact of mass media. London: Constable.

Siebert, F. S., Peterson, T., & Schramm, W. (1956). Four theories of the press: The authori-

tarian, libertarian, social responsibility and Soviet communist concepts of what the

press should be and do. Urbana: University of Illinois Press.

Strömbäck, J, Orsten, M, & Aalberg, T. (2008). Communicating politics: Political communi-

cation in the Nordic countries. Göteborg: Nordicom. van Dalen, A., de Vreese, C. H., & Albæk, E. (2012). Different roles, different content? A

four-country comparison of the role conceptions and reporting style of political jour-

nalists. Journalism, 13(7), 903–922. DOI: 10.1177/1464884911431538. van Kempen, H. (2007). Media-party parallelism and its effects: A cross-national comparative

study. Political Communication, 24, 303-320. doi: 10.1080/10584600701471674.

Voltmer, K. (2013). The media in transitional democracies. Cambridge: Polity

Hallin and Mancini Revisited 28

Endnotes

1 A thorough discussion of the advantages of typologies is provided by Bailey (1994).

He also mentions the pitfalls of this approach, such as the reification of ideal-types and

oversimplification (p. 11-16). These points also apply to comparative communication

research (Hallin & Mancini, 2012b). Abandoning the notion of media systems alto-

gether (Norris, 2009) would reduce our understanding of how different dimensions

and variables aggregate into bounded wholes, whose elements cannot be fully grasped

in isolation from another (Hallin & Mancini 20012b, p. 304).

2 Increased comprehensibility and reproducibility were the main reasons why we pre-

ferred the relatively popular cluster analysis over the less-established Qualitative Com-

parative Analysis (QCA) which is still very seldom employed for the purpose of typol-

ogy formation.

3 These coefficients may not completely meet the standards of scale development estab-

lished in micro-level research (DeVellis, 2003). However, our study can be considered

basic research and it is located on the macro-level where scholars frequently have to

cope with fewer cases and a considerable amount of noise in the data. So we had to

content ourselves, in the cases of two dimension indices, with levels of internal con-

sistency that are slightly lower than generally desirable but still acceptable (De Vellis,

2003, p. 95; Hair et al., 1998, p. 88). Low case numbers also prevented us from con-

ducting robust PCAs or EFAs in order to explore the dimensionality of our indices. So

we conducted PCA’s only as rough guidelines and just report their overall results. We

applied a stricter version of the Kaiser criterion, considering only components that

substantially exceeded an Eigenvalue of λ = 1 (DeVellis, 2003, p. 114). In this way,

each dimension could be condensed to one single component with Eigenvalues of λ >

1.4. These components each explained between 56 % and 78 % of total variance,

which can be considered satisfactory (Hair et al., 1998, p. 104).

Hallin and Mancini Revisited 29

Tables

Table 1: Original Dimensions and Models of Media Systems Model Mediterranean Northern European Dimension North Atlantic Polarized Democratic Liberal Pluralist Corporatist Press Market Low High High Political Parallelism High High Low Journalistic Professionalism Low High High Role of the State High High Low ES, FR, GR, AT, BE, CH, DE, CA, GB, Countries IT, PT DK, FI, NL, NO, SE IE, US Source: Hallin and Mancini (2004, p. 299)

Hallin and Mancini Revisited 30

Table 2: Data Sources Data Source Time Period Collection Method Population Sample Response Rate European Audi- 2011 Secondary collec- Public service Census Not available/ ovisual Obser- tion from Eurodata broadcasters Not applicable vatory (EAO) TV Worldwide and public service broadcaster’s an- nual reports European Elec- 2009 CATI Population aged 1000-1005 7-24 % tion Studies 18 and older per country (EES) I: Voter survey European Elec- 2009 Content Main national 2-4 TV chan- Not applicable tion Studies analysis television and nels and 3 (EES) II: Media newspapers newspapers study per country

European Media 2009-2010 Online survey Experts of me- 17-35 20-48 % Systems Survey dia and politics per country (EMSS) from academic institutions Hanretty (2009) 1944-2007 Data collection Public service 1-2 TV chan- Not applicable from PSB and Lex- broadcasters nel per coun- isNexis try

Humphreys 1996 Literature Public broad- Census Not applicable (1996) review casting systems Worlds of Jour- 2007-2011 Telephone and Professional 100 Not available nalism Survey face-to-face inter- journalists per country (WJS) views World Press 2008/2010 Secondary collec- Not applicable Not applicable Not applicable Trends (WPT) tion from national newspapers associ- ations and public institutions World Values 2005-2007 Face-to-face (and Population aged 657-2064 26-93 % Survey (WVS) telephone b) inter- 18 and older per country views b for Norway

Hallin and Mancini Revisited 31

Table 3: Operationalization of Dimensions Dimension Indicator Measure Data Transformation Scale Source Standard measures of national Overall daily WPT, market research institutes % newspaper reach 2008/10 (e.g. TNS Gallup) “People use different sources to learn what is going on in their country and the world. For each Filtering cases/re- of the following sources, please spondents describing WVS, % indicate whether you used it last themselves as 2005-07 week or did not use it last week “working class” to obtain information: […] Daily Working class newspaper” b Press daily newspaper Market a reach Filtering cases/ respondents de-scrib- “In a typical week, how many ing themselves as EES, % days do you follow the news?” c d “working class” and 2009 those who followed the news daily

Standard measures of national Women daily WPT, market research institutes % newspaper reach 2008/10 (e.g. TNS Gallup) a Average index of the three respective indicator indices (Cronbach’s α = .73); b for Finland, Germany, Italy, Norway, Spain, Switzerland, and USA; c for Austria, Belgium, Denmark, France, Great Britain, Greece, Ireland, Netherlands, and Portugal. d We used the percentage of respondents that follows the news seven days a week which is as equivalent to daily newspaper use as possible.

Hallin and Mancini Revisited 32

Table 3 (continued): Operationalization of Dimensions Dimension Indicator Measure Data Transformation Scale Source Lacking separation Number of evaluative refer- EES, of news and com- N ences per news story 2009 mentary “How far is the political coverage of each of the following media outlets influenced by a Additive index of party or parties to whom it is measures 1 and 2 EMSS, 0 to 20 close?” (1) b (Popescu et al., 2010 “To what extent does each me- 2011) dia outlet advocate particular Partisan influence views and policies?” (2) b and policy advocacy “Please tell me on a scale from 1 to 5 how influential are politicians in your work” (1) c WJS, “Please tell me how important Additive index of 0 to 20 d 2007- it is to set the political agenda, measures 1 and 2 2011 to influence public opinion, and to advocate for social change in your work” (2) c “The political orientation of the Political orientation EMSS, most prominent journalists is 0 to 10 of journalists 2010 Political well-known to the public.” Parallelism a Index based on a re- “In a typical week, how many gression model at days do you watch/read the fol- party level in which lowing news programs/ news- measure 1 predicts Media-party-paral- papers?” (1) EES, measure 2. The R2 0 to 100 lelism 2009 “How probable is it that you for each party is will ever vote for the following weighted according parties on a scale from 0 to to its electoral size 10?” (2) (van Kempen, 2007) “To what extent does each media outlet present equally well EMSS, Political bias Inverted scale 0 to 10 the arguments of all sides in 2010 political debates?” Rate of CEO turnovers (1) e Inverted average in- Rate of government changes dex of measures 1 Hanretty 0 to 1 followed by CEO turnovers and 2 (Hanretty, , 2009 within six months (2) e 2009) PSB dependence Three-point scale Hum- Politicization of public broad- based on categories 0 to 1 g phreys, casting systems f by Humphreys 1996 (1996) Note: a Average index of the six respective z-standardized indicator indices (Cronbach’s α = .83); b for all countries other than Switzerland and the USA; c for Switzerland and the USA; d the five-point scale of WJS (2007-11) was rescaled to the twenty-two-point scale of Popescu et al.’s (2011) index; e for all countries other than Greece and the Netherlands; f for Greece and the Netherlands; g the scale based on Humphreys (1996) was rescaled to the scale of Hanretty’s (2009) index

Hallin and Mancini Revisited 33

Table 3 (continued): Operationalization of Dimensions Data Dimension Indicator Measure Trans-for- Scale Source mation “And how much is the political coverage Inverted EMSS, in the following media outlets influenced 0 to 10 scale 2010 by its owners?” b “I have a lot of control over the work that WJS, 0 to 10 d I do.” c 2007-11 Internal “I am allowed to take part in decisions WJS 0 to 10 d autonomy that affect my work.” c 2007-11 “Some of these limits can come from within the news organization. Please tell Inverted WJS, me on a scale of 1 to 5 how influential 0 to 10 d scale 2007-11 each of the following is in your day‐to‐ day job.” c “Politicians, business people and interest groups influence what the news media re- Inverted EMSS, 0 to 10 port and how by pressurizing and bribing scale 2010 individual journalists.” b External “Other influences may come from outside autonomy the news organization. Again, please tell Inverted WJS, me on a scale of 1 to 5 how influential 0 to 10 d scale 2007-11 each of the following is in your work […].” c Journalistic “Journalists agree on the criteria for judg- Professionalism a EMSS, ing excellence in their profession regard- 0 to 10 2010 less of their political orientations.” b Professional “There are ethical principles which are so guidelines important that they should be followed by WJS, 0 to 10 d all journalists, regardless of situation and 2007-11 context.” c EMSS, “News media enjoy a lot of credibility.” b 0 to 10 2010 Media “Please tell me on a scale of 1 to 5 how credibility WJS, much you personally trust each of the fol- 0 to 10 d 2007-11 lowing institutions: The news media” c “Journalists are motivated by an ethic of EMSS, 0 to 10 serving the public interest.” b 2010 “Please tell me on a scale of 1 to 5 how important each of these things in your Public work is: 1) To provide citizens with the orientation information they need to make political WJS, 0 to 10 d decisions; 2) To provide the audience with 2007-11 the information that is most interesting; 3) To motivate people to participate in civic activity and political discussion.” c Note: a Average index of the five indicator indices (Cronbach’s α = .91); b for all countries other than Switzerland and the USA; c for Switzerland and the USA; d the five-point scale of WJS (2007-11) was rescaled to the eleven- point scale of EMSS (2010)

Hallin and Mancini Revisited 34

Table 3 (continued): Operationalization of Dimensions Dimension Indicator Measure Scale Source Market share of pub- Average daily market share % EAO, 2011d lic TV Public Public revenue (li- Broadcasting a Public revenue (USD) divided by cense fees) of public N EAO, 2011d GDP (USD) broadcasting TV ownership regu- Binary WPT, 2010 lation Newspaper/publisher Ownership Binary WPT, 2010 ownership regulation Regulation b Crossmedia (print/broadcast) Binary WPT, 2010 ownership regulation Press subsidies (USD) divided by Direct subsidies N WPT, 2010 GDP (USD) Press Subsidies c General VAT rate minus average Tax reduction press VAT rate (VAT single copy Percentage points WPT, 2010 and VAT subscription sales) Note: a Average index of the two respective z-standardized indicator indices (Cronbach’s α = .72); b average index of the three respective indicator indices (Cronbach’s α = .59); c average index of the two respective z-standardized indicator indices (Cronbach’s α = .60); d for the USA: CPB, 2009

Hallin and Mancini Revisited 35

Table 4: Increase of Heterogeneity by Agglomeration of Clusters Sum of Squared Change in Sum of Agglomeration Stage Number of Clusters Distances Squares (Δ SS) … … … … 7 10 7.6 2.2 8 9 9.9 2.3 9 8 13.7 3.8 10 7 18.3 4.6 11 6 23.2 4.9 12 5 30.1 6.9 13 4 37.3 7.2 14 3 49.5 12.2 15 2 66.3 16.8 16 1 96.0 29.7 Note: While agglomerating N > 4 clusters results in small changes of heterogeneity (Δ SS ≤ 7.2), merging cluster 3 and 4 increases heterogeneity significantly more (Δ SS = 12.2).

Table 5: Principal Component Analysis of the Dimension “Role of the State” Component Item Ownership Press Public Regulation Subsidies Broadcasting Crossmedia ownership .91 Press ownership .78 Tax reduction .83 Press subsidies .69 TV ownership .42 .49 Public revenue of public broadcasting .59 .78 Market share of public TV .77 Note: PCA with Oblimin rotation; N = 15; factor loadings α < .4 suppressed

Hallin and Mancini Revisited 36

Table 6: Correlations between the Dimensions Political Journalistic Public Ownership Press Dimension Parallelism Professionalism Broadcasting Regulation a Subsidies a Press Market -.58* .59* .43+ -.22 .21 Political Parallelism -.87** -.36 .24 -.15 Journalistic .21 -.34 .16 Professionalism Public Broadcasting -.05 .18 Ownership Regulation -.10 Note: N = 17; values are Pearson’s correlation coefficients; marked values are (or tend to be) statistically significant (+p < .1, *p < .05, **p < .01); a Correlation coefficients for these dimensions have to be treated with particular caution because the value distributions may not be suf- ficiently normal (Shapiro and Wilk’s W-test with p < .1)

Table 7: Comparison of Original Models and Empirical Types Original Model Empirical Type Democratic Corporatist Liberal Polarized Pluralist Northern DK, FI, NO, SE Central AT, DE, CH GB Western BE, NL IE, US PT Southern ES, FR, GR, IT

Table 8: Explained Variance by Original Models and Empirical Types Original Models Dimension Empirical Types Δ η2 (Hallin & Mancini, 2004) Press Market .36* .57* .21 Political Parallelism .62** .83*** .24 Journalistic Professionalism .61** .72** .11 Public Broadcasting .29+ .56* .27 Ownership Regulation .04 .62** .58 Press Subsidies .09 .42+ .34 Mean .33 .62 .29 Note: Values are partial η2 from separate one-way analyses of variance; marked values are (or tend to be) statistically significant (+p < .1, *p < .05, **p < .01, ***p < .001).

Hallin and Mancini Revisited 37

Table 9: Dimension Index Values and Country Rankings Dimension Political Journalistic Public Ownership Press Country Press Market Parallelism Professionalism Broadcasting Regulation Subsidies M Rank M Rank M Rank M Rank M Rank M Rank Austria -0.09 7 0.35 5 -0.90 14 0.36 7 1.11 1 -0.35 9 Belgium -0.74 15 -0.70 14 0.33 8 -0.42 13 0.06 7 0.52 5 Denmark -0.50 12 -1.32 16 1.31 1 1.41 1 -0.99 13 0.81 4 Finland 1.40 2 -1.36 17 0.92 4 0.77 4 -2.04 17 -0.73 15 France -0.99 16 0.62 4 -0.45 12 -0.02 10 1.11 1 0.47 6 Germany 0.96 5 -0.56 13 0.21 9 1.37 2 1.11 1 -0.66 12 Great Britain -0.17 8 -0.29 10 -0.49 13 0.89 3 1.11 1 -0.02 7 Greece -2.09 17 1.43 3 -1.35 15 -1.15 16 1.11 1 -0.39 10 Ireland -0.62 13 -0.06 7 -0.27 10 -0.49 14 -0.99 13 -1.08 16 Italy -0.33 9 1.74 2 -2.00 17 -0.29 11 0.06 7 0.87 3 Netherlands -0.33 10 0.19 6 0.72 6 -0.30 12 -0.99 13 -0.55 11 Norway 1.28 3 -1.04 15 0.87 5 0.50 5 -0.46 11 1.98 2 Portugal 0.38 6 -0.08 8 -0.36 11 -0.60 15 0.06 7 -0.32 8 Spain -0.62 14 2.10 1 -1.37 16 0.19 8 -0.99 13 -0.66 13 Sweden 1.61 1 -0.39 11 1.17 2 0.14 9 -0.46 11 2.14 1 Switzerland 1.21 4 -0.50 12 1.06 3 0.44 6 1.11 1 -1.37 17 USA -0.38 11 -0.12 9 0.60 7 -2.80 17 0.06 7 -0.67 14 Cluster Central (AT, CH, DE, GB) 0.48 2 -0.25 3 -0.03 3 0.77 1 1.11 1 -0.60 4 Northern (DK, FI, NO, SE) 0.95 1 -1.03 4 1.07 1 0.71 2 -0.99 4 1.05 1 Southern (ES, FR, GR, IT) -1.01 4 1.47 1 -1.29 4 -0.32 3 0.33 2 0.07 2 Western (BE, IE, NL, PT, US) -0.34 3 -0.15 2 0.20 2 -0.92 4 -0.36 3 -0.42 3 Note: Values are z-standardized dimension indices.

Hallin and Mancini Revisited 38

Figures

Figure 1: Hierarchical Clustering Dendogram

Northern

Southern

Western

Central

Figure 2: Cluster Profiles

Note: Values are z-standardized dimension indices.